Varnhagen.pdf

726 C. K. Varnhagen et al.
misspelling. Neither of these categories is an example of new language so we
classified these separately as Errors.
Some words received more than one classification; for example, from Table 1,
the insider word, hottie, was in all uppercase letters and so was scored both an
insider word and as upper case. Other examples of multiple classifications include
im, classified as lowercase and as a contraction error, and lol, classified both as a
short cut acronym and as a pragmatic device emotion acronym. Approximately 3%
of the words were scored as more than one category.
The conversations were initially scored by a group of researchers (the authors and
others) to define the categories of new language. The conversations were then
rescored by researchers naive to the initial discussions, using the categories of
definitions. A 20% sample was then scored a third time by a new naive rater.
Agreement on overall identification of a word a new language was 99%. Cohen’s
kappa (Cohen, 1960) for inter-rater agreement on scoring into the different
categories was j = 0.97, indicating excellent inter-rater reliability and ease of
classifying instances of new language according to our classification scheme.
Results
As this was an exploratory study and we conducted a large number of analyses, we
a priori guarded against family-wise error by applying Bonferoni corrections
(dividing alpha = 0.05 by the number of tests) to each set of statistical analyses
conducted. Only results significant beyond the corrected alpha level are reported
below.
New language use
Participants used an average of 31.8 instances of new language (SD = 14.1,
range = 7–89) in their instant messages. Table 2 shows the mean frequencies in
new language use for the different categories of new language. The means among
the different categories add up to greater than the overall total because a few words
could be classified into two categories (e.g., im is classified as both lower case and a
contraction error; WUT is classified as both phonetic and upper case).
Considering first the three groupings of categories, short cuts, pragmatic devices,
and errors, we conducted a single factor repeated measures analysis of variance with
the three general categories of short cuts, pragmatic devices, and errors as the repeated
measures factor and participants as the random factor. We found a statistically
significant difference among these three general categories, F(2, 78) = 53.07,
p \ 0.001. Participants most often used short cuts in their instant messaging, followed
by pragmatic devices to support their messages, followed by errors, Tukey’s post hoc
HSD = 5.24, p \ 0.01.
We also analyzed use of new language in each of the three general categories.
We conducted a single factor repeated measures analysis of variance with the eight
different shorts cuts categories as the repeated measures factor and participants as
the random factor. We found a statistically significant difference among these three
123